Look for the pile no one is assigned to on any high-volume role: applications that plainly aren’t right for the role and haven’t had a reply. The recruiter is busy at the promising end of the inbox, so that pile only grows, and every person in it is still waiting to hear.

AI candidate screening gets every application read and given a fit label, however many arrive. That pile stays where it is, though, because a fit label is information for a recruiter, and an application leaves the pile only when a recruiter decides its next step and someone sends the reply.

So I’d build the playbook after the sort, where every application goes down one of three routes with owners, each with a reply time the candidate can count on, and the middle route stays generous until the recruiters’ own record says it can shrink.

Why the playbook needs only three routes.

Once an application has its fit label, a recruiter has three moves open, and the playbook’s three buckets are those moves, so from here on I call them routes: advance, recruiter reads, and decline with care.

A fourth route, the “maybe later” or “hold for the hiring manager” kind, rebuilds the unowned pile from the opening. No one has the job of emptying it, so the people in it wait longest of all.

A screening tool may label fit in four or five steps, and the count of routes stays at three. The labels describe applications and the routes describe what recruiters do, so a recruiter can send applications with different labels down the same route.

Put each route on paper with an owner and a clock.

Agree the three routes before the role opens, with the hiring manager in the room, because the decline route rests on must-haves they signed off.

Publish the reply times where the hiring manager sees them as well, at the top of the role’s intake notes or in the hiring channel. Candidates live by that column, and a reply time only the recruiter can see is the first thing to slip in a busy week.

A must-have written as an adjective, such as “great attention to detail,” can’t be checked against a page. This guide to AI resume screening shows how to rewrite one as evidence an application can carry.

Send the middle to a recruiter, then let it shrink.

In the first weeks of a role, I’d send every application without an obvious route to a recruiter’s full read. It’s more reading than anyone wants, and it shows you where the fit labels are right for this role before a label alone is enough to send someone toward a call.

For each application that leaves the middle, note the route the recruiter chose and the line in the application that settled it. Those notes are the only evidence I’d trust for moving a boundary between routes.

If nearly everything the recruiter reads goes on to advance, the advance route is too strict, so widen it. If nearly everything goes to decline over the same missing must-have, write that must-have into the profile your screening tool reads against, so its labels reflect it before a recruiter has to. Then leave the route rules alone for a week, because two changes at once leave the record unreadable.

Read the middle against the must-haves you wrote, too. When an application keeps landing there because two requirements pull against each other, this piece on settling conflicting signals sets out a rule for each kind of clash.

Close the decline route the same week.

The decline route is the easiest one to let slip, because no one on the team is waiting for it, while the candidate is. Assign it an owner and a reply time like the other two, and have that recruiter confirm each decline before the reply goes.

With a narrow decline rule, confirming takes moments, because the recruiter only checks that the reason names a missing must-have. A reason resting on a preference, or on a weak overall fit, sends the application to the middle route instead, since the hiring manager agreed the decline route for missing must-haves only.

If the person looks right for another open role, tag them in your applicant tracking system before the reply goes, so the recruiter on that role can find them. Then send a short, polite reply that says plainly this role is closed to them.

Carry the open question into the first call.

If an application’s reason leaves something the role needs unclear, make that the first question you ask on the call it advances to.

If the reason says the resume shows customer-facing work and no team the person ran, ask about the team before anything else. Write the answer down against the line of the profile it settles, so the next interviewer starts from it instead of asking again.

Where Metaview fits in AI candidate screening.

Every route above starts with a recruiter reading why an application looks the way it does, and Application Review is where that reason comes from. For each role, Application Review holds each application up to the ideal candidate profile (ICP) a recruiter approved and sorts it into one of four ICP Fit buckets: Great, Good, Okay, or Poor. The bucket is a label rather than a score, and its reasoning is shown beside it.

The routes stay a recruiter’s work, application by application. In a role’s first weeks, I’d have the recruiter take the four ICP Fit buckets this way:

• Great Fit: the recruiter checks each one against the must-haves before booking the call.
• Good Fit: the recruiter reads it in full, and it’s the first bucket I’d let go straight toward a call once the record supports that.
• Okay Fit: the recruiter reads it in full.
• Poor Fit: the recruiter reads the reasoning for a named missing must-have and declines only when one is there. Anything else gets a full read.

Application Review doesn’t auto-reject anyone, and the accept or reject call on each application stays with the recruiter, so every decline is a person’s, made on that application.

When the middle route’s record keeps pointing the same way, look at the ICP behind the buckets. Application Review suggests changes to the ICP as your recruiters accept and reject, and the record tells the recruiter which suggestion is worth approving. After a recruiter approves one, Application Review re-evaluates the whole list against the edited ICP. Before any of that, a recruiter approved the ICP in the first place, after reviewing the draft Application Review produced from the job post and whatever extra context the team supplied.

When an application arrives, Application Review also checks it for identity deception and for application automation, and anything flagged shows the recruiter a risk level with a plain-language explanation. I’d treat a flag as a reason for a full read and never as grounds for a decline by itself.

Workleap, a people management platform, turned to Application Review when the applications for each role were arriving by the hundreds within days. Its recruiters can now “quickly identify which candidates to move forward with, which to stop, and those who might deserve a closer look.” Senior Recruiter Johnny Drexhage describes the speed: “Both strong and weak profiles are reviewed within a couple of seconds.”

When Application Review launched, Metaview’s chief executive described where the tool’s work ends and the recruiter’s begins.

The open question from an application gets its answer on the first call. If the people on that call agree to recording under your team’s consent steps, the Metaview Notetaker takes part as a visible participant, and the call comes back recorded and transcribed, then set out as structured notes.

Metaview captures every spoken word from both sides of the conversation. The recruiter choosing the candidate’s next route can then read the answer in the candidate’s own words rather than reconstruct it from memory.

If the middle route keeps raising the same question across one role, such as notice period, a recruiter can put it to Metaview Reports, which answers from interview data your team recorded.

1

1. 1The page’s line about pulling out any data point mentioned in conversations, which is where a question the middle route keeps raising would go.

Time one day of applications to their replies.

Pick one day of applications from the role’s first week, and at the week’s end look at how many working days each one waited for its reply, whatever route it took. If the declines waited longest, the decline route has no owner in practice, and that gap gets closed before anyone touches the screening setup.

If every reply landed inside the times you published, the routes are holding, and the middle route can start to shrink on the record the recruiters kept.